College betting model

[Bill the cop]

"But like I said I dont like to comment on certain things."

I probably should have had the same mindset.

For the record, we use Don Best closers in our database.

Just for what it's worth, of the 719 games last year (regardless of the closing lines or LTR), NOT ONE GAME HAD A 25-0 or 26-0 score at the half.

[Geor99]

MonkeyF0cker Can you expand on negating my hypothesis with more than "no?" I don't want to argue, but I would like to see your point. If the spread and total are the same, wouldn't all of the scoring potential (for lack of a better term) be in the favorite? If it is Fav-27 with a total of 27, wouldn't this be the case. It has to. There is no room in the total for the dog to score if the favorite is expected to score all of the points. The market is assuming 0 production by the dog and 100% of the production by the favorite.

Please comment, I always like to learn. My explanation above is how I understand the market.

[donjuan]

The problem with people who THINK they know something and post stupid stuff like you do, is that they just dig themselves deeper and deeper. It's obvious you don't have any type of database or you wouldn't have made that statement.

Over the last 3 years there have been 139 games that met the 2:1 LTR (which are the games we're talking about). Do you know how many of those games had either team score exactly 25 or 26 points in the first half? Of course you don't. Well let me enlighten you, NONE. That's right zero, nada!

I'm not saying it COULDN'T happen, I'm just saying it would be an unrealistic football score for one team to score exactly 25 or 26 points in the 1st half, and even rarer for the score to be exactly 25-0 or 26-0.

Having an understanding of the market >>>>>>>>>>>>> having a database used strictly for data mining

[tomcowley]

MonkeyF0cker Can you expand on negating my hypothesis with more than "no?" I don't want to argue, but I would like to see your point. If the spread and total are the same, wouldn't all of the scoring potential (for lack of a better term) be in the favorite? If it is Fav-27 with a total of 27, wouldn't this be the case. It has to. There is no room in the total for the dog to score if the favorite is expected to score all of the points. The market is assuming 0 production by the dog and 100% of the production by the favorite.

Please comment, I always like to learn. My explanation above is how I understand the market.

It's functionally true in something like NCAAFB, but the dog can theoretically score as long as the distributions are still pure. In billthedonk land, where team scores of 25 and 26 are impossible, the dog could safely get a 1-point safety on a line of -25.5/o25.5 and never blow the FO/DU purity.

[wantitall4moi]

Having an understanding of the market >>>>>>>>>>>>> having a database used strictly for data mining

having a database with openers, closers, line moves (timestamped) along with vig from about a dozen books >>>>> than anything. Well the only thing better would be having the books accounting sheets.

because then you can look at what the 'market' did and how it was right or wrong and if the closer was indicative of what the 'market' price truly was.

But it still doesnt make it foolproof there is still a margin for error. You can show 5 guys the exact same stuff, let them use your info and they will all come up with 5 different ways to interpret it. And all 5 could be equally valid.

the 'market' still originates with that opening line, and no matter how far off it is (according to people who think they know how it works) it can still only move so much to avoid a catastrophe. If a book posts a team -140, but someone has a model that shows they should be +120 you think books would move that much? Not frigging hardly. the +120 might even be more valid. But unless the +130 side goes crushed the books will stick to what they have. Unless they hear a syndicate or something might hammer it then they might move some on air, but they are still limited in how far they can move it.

But generally the numbers arent nearly that inefficient. They do what theyre supposed to do, get as close to balanced action as they can or at least enough to allow the book to make a little without much risk.

[MonkeyF0cker]

MonkeyF0cker Can you expand on negating my hypothesis with more than "no?" I don't want to argue, but I would like to see your point. If the spread and total are the same, wouldn't all of the scoring potential (for lack of a better term) be in the favorite? If it is Fav-27 with a total of 27, wouldn't this be the case. It has to. There is no room in the total for the dog to score if the favorite is expected to score all of the points. The market is assuming 0 production by the dog and 100% of the production by the favorite.

Please comment, I always like to learn. My explanation above is how I understand the market.

Assuming that the market is unbiased, the listed spread and total are simply medians of the scoring distribution. With that line, there is vastly higher potential for scoring in the favorite. However, there is certainly still potential for the underdog to score. Albeit, at a significantly lower probability...

[Bill the cop]

Having an understanding of the market >>>>>>>>>>>>> having a database used strictly for data mining

Gee, "Having that understanding of the market" must really give you a feel for how often a 1st half ends 25-0 or 26-0. Afterall, off the top of your head, think about Boise, Oregon, and USC, hell they go for 2 all the time, right?

Now, lets look at FACTS and actually SEE how often 1st halfs end 25-0 or 26-0:

Last 4 years, sample size 2862, games that ended 1st half 25-0 or 26-0, exactly ONE (11/14/2009 UCLA 26, Wash St 0)

So when I said it was a rare occurence that was a big understatement, yeah, those databases are pretty useless, (unless you want facts).

[MonkeyF0cker]

Gee, "Having that understanding of the market" must really give you a feel for how often a 1st half ends 25-0 or 26-0. Afterall, off the top of your head, think about Boise, Oregon, and USC, hell they go for 2 all the time, right?

Now, lets look at FACTS and actually SEE how often 1st halfs end 25-0 or 26-0:

Last 4 years, sample size 2862, games that ended 1st half 25-0 or 26-0, exactly ONE (11/14/2009 UCLA 26, Wash St 0)

So when I said it was a rare occurence that was a big understatement, yeah, those databases are pretty useless, (unless you want facts).

And how many first halves were lined ±24 with a total of 28?

[donjuan]

Assuming that the market is unbiased, the listed spread and total are simply medians of the scoring distribution. With that line, there is vastly higher potential for scoring in the favorite. However, there is certainly still potential for the underdog to score. Albeit, at a significantly lower probability...

The way I understand this, and correct me if I'm wrong, is that the dog can only score when the favorite has an extremely high or low score. Logically they could have a 0% chance of scoring against the first string defense and the opposing team only pulls the first string defense when they are up by x amount of points with y time left such that they have a non-zero chance of scoring but that they do still have a 0% chance of covering.

[donjuan]

Gee, "Having that understanding of the market" must really give you a feel for how often a 1st half ends 25-0 or 26-0. Afterall, off the top of your head, think about Boise, Oregon, and USC, hell they go for 2 all the time, right?

Now, lets look at FACTS and actually SEE how often 1st halfs end 25-0 or 26-0:

Last 4 years, sample size 2862, games that ended 1st half 25-0 or 26-0, exactly ONE (11/14/2009 UCLA 26, Wash St 0)

So when I said it was a rare occurence that was a big understatement, yeah, those databases are pretty useless, (unless you want facts).

The game of football is constantly evolving. Offenses are vastly different today than they were 15 years ago. Coaches change their decision making. Rules change. If you think your data from 15 years ago is better than the methods the market uses to set lines we can all have a big lolfest.

Edit for the simpletons: I'm talking about BTCs data in general. In this case it's hilarious to include pk games with totals of 20 and the like when analyzing scores of 25-0 and 26-0.

[MonkeyF0cker]

The way I understand this, and correct me if I'm wrong, is that the dog can only score when the favorite has an extremely high or low score. Logically they could have a 0% chance of scoring against the first string defense and the opposing team only pulls the first string defense when they are up by x amount of points with y time left such that they have a non-zero chance of scoring but that they do still have a 0% chance of covering.

Take it in terms of a significant sample of Monte Carlo simulations, just as a generalization (since I don't have an actual distribution to work from), your sample would likely contain a lot of results akin to 21-6, 24-3, 27-0, 28-0, 31-0, 31-3, etc. including a share of less conventional scores. And you'll even have your outliers of 14-14 or 49-6. There is certainly a chance that the underdog scores. However, the median of the total and the median of the spread happen to converge on the same number.

[donjuan]

Take it in terms of a significant sample of Monte Carlo simulations, just as a generalization (since I don't have an actual distribution to work from), your sample would likely contain a lot of results akin to 21-6, 24-3, 27-0, 28-0, 31-0, 31-3, etc. including a share of less conventional scores. And you'll even have your outliers of 14-14 or 49-6. There is certainly a chance that the underdog scores. However, the median of the total and the median of the spread happen to converge on the same number.

In that case I'm not sure I understand then. If there are 21-6s and 24-3s in there, the favorite loses but the over pushes. What scenario makes up for that on the other side since you can't have negative scores and any time the favorite wins by more than 27 the game goes over also?

[MonkeyF0cker]

In that case I'm not sure I understand then. If there are 21-6s and 24-3s in there, the favorite loses but the over pushes. What scenario makes up for that on the other side since you can't have negative scores and any time the favorite wins by more than 27 the game goes over also?

Frequency makes up for it. Those scores would be far less likely than something like 31-0, 28-0, 27-0, etc. However, the underdog still has a chance of scoring.

[tomcowley]

Frequency makes up for it. Those scores would be far less likely than something like 31-0, 28-0, 27-0, etc. However, the underdog still has a chance of scoring.

Let's just put it on half points for simplicity. -25.5/o25.5. If these lines are fair, F, O, D, U, all have to hit 50% individually. Since every F cover is automatically an O, that means FO has to be .5 (and FU 0 by the lines). Then by efficiency, DU has to be .5, and FU/DO 0. So you can't ever have a DO outcome with these lines, so the only way the dog can score is if the fav has a scoring hole (like in my post 144) that prevents the dog score from ever resulting in a DO.

[MonkeyF0cker]

Let's just put it on half points for simplicity. -25.5/o25.5. If these lines are fair, F, O, D, U, all have to hit 50% individually. Since every F cover is automatically an O, that means FO has to be .5 (and FU 0 by the lines). Then by efficiency, DU has to be .5, and FU/DO 0. So you can't ever have a DO outcome with these lines, so the only way the dog can score is if the fav has a scoring hole (like in my post 144) that prevents the dog score from ever resulting in a DO.

Yes. That's a good way to put it, Tom.

[wiffle]

funkym0cker4life

thanks for the royals

100 dimes baby, no joke

[Justin7]

I talked to BTC for awhile. I think he observed something interesting, but perhaps didn't explain it in the way pro bettors would expect.

Observations:
1. In the past, the 1h spread for an NCAAF favorite is around 60% of the game spread. The total is about 50% of the game total.
2. The sportsbook conversions from #1 are generally accurate, but not when the LTR ratio is extreme (which occurs with huge spreads, and surpisingly at games near Pick with no medium favorite).
3. When the LTR ratio is at either edge, you are seeing a long-standing market inefficiency in h1 lines. The conversion in #1 is wrong (except for "average" games), and the market has not adapted.
4. Until the market to adapts (and it will, it is only a question of when), expect to see BTC's h1 strategy of 2 straight bets on certain NCAAF games win.
5. BTC's observations about the negative correlation of certain subsets of h1 bets are interesting, but not nearly as important as #3.

Conclusions:
1. BTC's approach is likely to make some money this year, but there are better ways to exploit what he has observed.
2. The market will correct itself (in no small part due to this thread), and this opportunity will disappear when sportsbooks develop better conversion charts.

[Dark Horse]

to add to wehat the guy above said. People have been raised on tout speak for the most part. Used to be that 'information' was the key ingredient to beat the line. 'We have inside info', heard it all the time, still hear it today. Ten once the guys figured out the 'inside info' wasnt enough to sell themselves they started with 'data' and stats and trends, They all started getting databases and could look up things in the blink of an eye, they found stuff that had some subsets or appeared to and twisted and used that to sell to people. As it progressed and they could search more parameters they created 'super systems' where they would pull shit like 20-0 over the past 10 years or 31-2 the last 6 seasons. We have all seen them so we know the pitch.

Then once people figured that those were also useless they started in with trying to validate themselves with economic talk. They started preaching bank roll and money management, then the sports betting world was like the stock market and they knew the market better than anyone, they had inside dope with boos and knew what side was heavy and what side wasnt, they knew what side the 'sharp' guys were betting, they knew what lines were getting bet and not moving, and thus were being gambled on by the books. This is the era we are in now. The market era. Once people figure out that that is bullshit too they will come up with another make believe catch phrase.

While 'market' in a generic term is true it isnt what people try to make it out to be. But as it pertains to this debate everyone thinks theyre an expert and thus eevryone thinks they can contribute. The irony is theyre all probably half right. i didnt read it, but I am sure whatever numbers they have are valid, they just dont realize their numbers arent the only ones. So when they try and determine something they arent using all the facts. If you have 6 books you can have 6 different offerings for the same game. Lines might be same but vig could be different, or lines might be different but with similar vigs. Any multitude of factors, and all those factors over all the examples will create a mish-mash of results. The 'purest' way to do it is use a single book with their odds and their results and bet with them exclusively. That is really the only way to control it. But if you use several sources,and I am not talking Stu Feiner, you are going to corrupt the data. Probably wont invalidate it, but if you dont get the correct side of it it will be useless.

Thats why these debates rage on and on between math guys, handicappers, stats guys you name it. They all have some validity, but none of them really look at everything. But when you do you realize a lot of things.

One final thing about subsets. With the right database you can find alot, probably hundreds, that have 80% plus returns in football or basketball. The trick is to make sure you dont reduce a subset so much it invalidates it. If the generic subset is 60% with 200 results, and you dont think that is good enough so you keep adding criteria until you get something that is 20-3, you have over analyzed it. The original of say 128-83 was probably as far as you needed to go. But in everyones mind (just about) 86% is better than 60%.

But I always laugh at these threads with supposed smart guys in them, a person is spoon feeding you something to make you money (which is really kind of dumb in the first place) and no one wants to take advantage of it. Then when the books change the rules those same dopes are the ones complaining about books killing all the +ev bets. Its sad and comical at the same time.

Much truth here. A rare guy who isn't afraid to think outside of the box. Most of these math guys are completely in-the-box, like Jack's little babies, and they don't know it. They just love shaking their rattlers like pharaohs scepters.

[dogman]

As with the old correlated parlays-F/O and D/U which have dissappeared as they became well known and talked about on the forums so too will these(D/O). I would love to see this thread go away as another good correlation will be a thing of the past as it too becomes more talked about.

The same thing happened with teasers as most books have either changed the lines to make it harder to play them or changed the payoffs to make it more difficult to win.

[wantitall4moi]

Conclusions:
1. BTC's approach is likely to make some money this year, but there are better ways to exploit what he has observed.
2. The market will correct itself (in no small part due to this thread), and this opportunity will disappear when sportsbooks develop better conversion charts.

This is the most important thing. While it is nice to talk about this stuff and try and sound smarter than the average guy, or get into a pissing contest with another smart guy (who may very well be trying to debunk it on purpose) it is best to leave this stuff to yourself or a friend or two. There ARE still a very few things that can be worked with. Some people know some of them, but some of them they dont. They also dont bet them in the most efficient manner either. They bet them in the basic "this can win no matter what' manner. So books might not notice so quickly.

Books are in the same myopia a lot of bettors are. or at least the math guys. They find a model and it works so they use it. Until someone beats it, then the book with the most smarts changes and then all the other ones follow suit. (and most of those books dont even know why they changed they just know the other book is better than they are)

That is why I like to speak in generalities or in circular ways. I still try to get my point across but in some cases, people lack comprehension I guess.

Bottom line is it is better to find one of these and use it as long as you can. Than it is to go on forums and try to impress people, especially for free. But that is the nature of the beast. People like to be repected or seen as smart or intuitive or whatever. But the truly smart guys are the ones making money and not closing advantages, and being liked or respected for their overall knowledge not a knowledge or one specific thing. I really dont care what people thin about me, and that might be the best attitude to have, it saves alt of typing trying to defend yourself. Make a point if someone disagrees make another point, if that isnt accepted then maybe a third one, if they dont get it by then so be it. But if it is a debate about something truly advantageous stop before you make it too obvious. Generally speaking if guys are worth their 'reputations' they should be able to look at something someone says (in enough general info is given) and proof it out for themselves. if they cant then they shouldnt be arguing about it. But that can still serve a purpose, it can dissuade some people who respect that guy from believing it and thus help the play stay under the radar.

[Bill the cop]

I talked to BTC for awhile. I think he observed something interesting, but perhaps didn't explain it in the way pro bettors would expect.

Observations:
1. In the past, the 1h spread for an NCAAF favorite is around 60% of the game spread. The total is about 50% of the game total.
2. The sportsbook conversions from #1 are generally accurate, but not when the LTR ratio is extreme (which occurs with huge spreads, and surpisingly at games near Pick with no medium favorite).
3. When the LTR ratio is at either edge, you are seeing a long-standing market inefficiency in h1 lines. The conversion in #1 is wrong (except for "average" games), and the market has not adapted.
4. Until the market to adapts (and it will, it is only a question of when), expect to see BTC's h1 strategy of 2 straight bets on certain NCAAF games win.
5. BTC's observations about the negative correlation of certain subsets of h1 bets are interesting, but not nearly as important as #3.

Conclusions:
1. BTC's approach is likely to make some money this year, but there are better ways to exploit what he has observed.
2. The market will correct itself (in no small part due to this thread), and this opportunity will disappear when sportsbooks develop better conversion charts.

Oh come on Justin, you know all my analysis is just "datamining garbage" and no "sharp" bettor would pay any attention to it, right?

In my defense of "spilling the beans", you are well aware of several things I do that never see the light of day on these forums. I'll just leave it there.

[MonkeyF0cker]

3. When the LTR ratio is at either edge, you are seeing a long-standing market inefficiency in h1 lines. The conversion in #1 is wrong (except for "average" games), and the market has not adapted.

Really? Why is nobody putting this in terms of standard deviation? Is it because it would essentially invalidate it? Or is it just ignorance?

4. Until the market to adapts (and it will, it is only a question of when), expect to see BTC's h1 strategy of 2 straight bets on certain NCAAF games win.

Uh huh. Ok.

5. BTC's observations about the negative correlation of certain subsets of h1 bets are interesting, but not nearly as important as #3.

Interesting if you enjoy reading useless information.

Conclusions:
1. BTC's approach is likely to make some money this year, but there are better ways to exploit what he has observed.

You have a crystal ball, Justin? Can you offer any insight into why this observation is statistically significant? No? Didn't think so.

2. The market will correct itself (in no small part due to this thread), and this opportunity will disappear when sportsbooks develop better conversion charts.

Yes. Everyone is watching this thread closely just waiting to pounce. Lead the sheep to slaughter.

[MonkeyF0cker]

Let's just put it on half points for simplicity. -25.5/o25.5. If these lines are fair, F, O, D, U, all have to hit 50% individually. Since every F cover is automatically an O, that means FO has to be .5 (and FU 0 by the lines). Then by efficiency, DU has to be .5, and FU/DO 0. So you can't ever have a DO outcome with these lines, so the only way the dog can score is if the fav has a scoring hole (like in my post 144) that prevents the dog score from ever resulting in a DO.

I should add that there may also be the possibility of some DO subsets but their probabilities are slight and are simply absorbed by vig.

[Justin7]

Really? Why is nobody putting this in terms of standard deviation? Is it because it would essentially invalidate it? Or is it just ignorance?

Would you really like me to spell out exactly what the sportsbooks are doing wrong, to what degree they are wrong, and what certainty? And give a methodology for evaluating that?

[MonkeyF0cker]

Would you really like me to spell out exactly what the sportsbooks are doing wrong, to what degree they are wrong, and what certainty? And give a methodology for evaluating that?

LOL. You wrote a fukking book attempting to do just that. Why the Hell wouldn't you?

The results can simply be attributed to variance. Period. Standard deviation. Look it up.

[GoIrish682]

help me out here

is BTC considered a 'math geek' or the other less tangible 'gambling thinking' type?

[Justin7]

LOL. You wrote a fukking book attempting to do just that. Why the Hell wouldn't you?

The results can simply be attributed to variance. Period. Standard deviation. Look it up.

Monkey,

You talk a lot about a book you never read.

[MonkeyF0cker]

From a thread entitled "Is Justin7 writing a sports betting book?"...

I'm working with a Ph.D. mathematician from Hungary to try to get my MLB runline conversion chart right. I'd never even heard of a "logistic regression" until he showed me a possible solution to estimating the odds of the home team winning by exactly one run, given a game moneyline and total.

Are you telling me that you didn't provide "your" conversion methodology in your book?

And why would I read your damn book, Justin? How often do you think chefs read the McDonald's cookbook?

[durito]

Lots of +ev run lines everyday at pinnacle based on his runline conversion chart. You should start betting them MF, you could join me in poverty soon.

[Zojo274]

Hey Guys if anyone still following this thread I have 4 games this weekend I THINK meet BTC's low LTR criteria and I intend on betting.
I am sure this thread is way past his first post but anyway 4 what its worth Peace and money to all.
#313 utep + 29 o/u 47.5
#319 E. Mich + 29 o/u 49
#379 Tulsa + 31.5 o/u 62.5
#385 FL Atl + 33 o/u 57.5
One bet on dog one bet on ovr for each

[Rambo]

You guys are making this too complicated, just pick the team with the coolest helmets.

[WendysRox]

Man, I'm a little slow, and I know this, but I've read 3 pages of this stuff and still don't understand what constitutes a play for you.

[Zojo274]

i understanding is
High spread game prerequisite


Total/spread = LTR
If LTR is < 2.0 bet the dog to cover and the over.
If you think through this you can see the logic see post above

[Peregrine Stoop]

i understanding is High spread game prerequisite Total/spread = LTR If LTR is < 2.0 bet the dog to cover and the over. If you think through this you can see the logic see post above

but, there is a very important caveat

it's only a play as long as the spread and totals for the first half stick to what they've historically been as a ratio compared to the full game. When the books adjust (and some have in the past, for instance BetPhoenix would shade these last year), there will be no more plays. It's a problem at the tails with the models being used.

[Zojo274]

Hey Guys if anyone still following this thread I have 4 games this weekend I THINK meet BTC's low LTR criteria and I intend on betting.
I am sure this thread is way past his first post but anyway 4 what its worth Peace and money to all.
#313 utep + 29 o/u 47.5 W/W
#319 E. Mich + 29 o/u 49 W/ L
#379 Tulsa + 31.5 o/u 62.5 W/L
#385 FL Atl + 33 o/u 57.5 W/L
One bet on dog one bet on ovr for each

5-3 OK I PLACED $500 ON EACH SO +$850 tRY AGAIN THIS WEEK